Absolute Pressure Control Boosts Oxide Uniformity

		At a Glance
	As gate oxide thicknesses decrease, controlling them becomes a bigger challenge. They are even sensitive to changes in barometric pressure. Controlling exhaust pressure variations can help control gate thicknesses.

As gate oxides begin to approach the theoretical minimum thickness, control of thickness uniformity is becoming both more important and more difficult. Much work has been done to improve control over all aspects of the process, including improved mass flow controllers for accurate gas control, advanced thermal control and relative pressure control to minimize the effects of house exhaust fluctuations.

While this work has made improvements in oxide uniformity, significant variation still exists, which recent evidence suggests may be due to changes in barometric pressure. Although absolute pressure control has long been a critical control parameter for deposition processes, until recently no one has attempted to control the absolute pressure in a thermal oxidation process.

A 1998 study made by SC300 engineers showed a high correlation between average thickness and barometric pressure for oxidations performed in an SVG vertical furnace. Figure 1 shows the average thickness plotted in red, with the ambient barometric pressure at the time of wafer processing shown in blue. As the pressure rises, the oxide tends to be thicker. As the pressure falls, the oxide shows a predictable thinning.

Figure 2 shows the same data as shown in Figure 1. To show the close link between barometric pressure and oxide thickness, each set of thickness data was plotted against a normalized barometric pressure. Plotted in this way, all the thickness points lie inside the overall range specification. Based on these results, the SC300 engineers began to request absolute pressure control for their furnaces.

How absolute pressure affects T_OX

More recent data by Martin Fallon (currently of National Semiconductor)¹ show similar effects of barometric pressure on a horizontal furnace (Fig. 3). Fallon suggests a linear relationship between film thickness and absolute tube pressure. To test the theory, Fallon created an absolute pressure control system by exhausting into an intermediate chamber and controlling the tube pressure by increasing or decreasing nitrogen flow into the intermediate chamber to compensate for atmospheric pressure changes. Figure 4 shows the influence of absolute pressure on oxide thickness in the case of controlled pressure. It is apparent from this chart that, within this control region, pressure does have a nearly linear effect on oxide thickness.

The scope of the problem

Altitude and weather combine to affect barometric pressure in any location at any given time. The effects of altitude are constant, based on the location's height above sea level. The effects of weather can change rapidly — dropping quickly when a storm front moves in or remaining high with fair weather.

The lowest barometric pressure ever recorded at sea level was 877 mbar (25.90 inches Hg), and the highest pressure recorded was 1185 mbar (35.01 inches Hg).² This represents a swing of more than nine inches Hg. For those more familiar with house exhaust pressures, this swing equates to 123.9 inches (2787 mm) of water. For those involved in low-pressure processing, nine inches of mercury is equivalent to 203 Torr.

This large range in atmospheric pressure is due to weather. Altitude also contributes to barometric pressure differences. In Colorado Springs, Colo., where the altitude is about 6135 feet, the mean barometric pressure is 811 mbar (23.9 inches Hg). At sea level, the barometric pressure on a nice day is around 1013 mbar (29.9 inches Hg).² In a single location, storms can significantly reduce the air pressure. While the range of a storm's effect is less than that of altitude, the speed of the change is more significant. A typical storm may cause a pressure change of 33 mbar or an inch Hg within an hour. Hurricanes and monsoons can cause changes of well over 60 mbar (2 inches Hg) during their passage, with an equally rapid recovery.

4. After adding pressure control to a horizontal furnace, changes in absolute pressure set point showed a clear relationship between oxide thickness and barometric pressure.

To put this into perspective, a low-pressure storm front could cause a change of 13 inches (330 mm) of water or 24.3 Torr in the process pressure² of a thermal oxidation process running with relative pressure control. This is equivalent to a 3% fluctuation in absolute pressure. Whether this change occurred during a single run or occurred between runs, the result would be the same: about a 3% change in oxide thickness.

Controlling absolute pressure

A new pressure controller promises to reduce these pressure-related oxide thickness fluctuations. The controller is based on existing relative pressure control technology (see sidebar) with the addition of a sophisticated atmospheric pressure sensor and a proprietary pressure-reducing technique. The controller has a control range from 800 to 1100 mbar, spanning almost the entire range of earthbound pressures. This allows users to program the controller (either mechanically or through the software) for the average barometric pressure for their area. This strategy would minimize the change on an existing process, for example. Alternatively, the user can adjust above or below the average pressure as required.

5. Set point changes using the new absolute pressure controller create easily measured pressure changes in the process chamber (a 300 mm RVP furnace).

Figure 5 shows the results of a measurement test using this new pressure controller. The graph shows measured tube pressure with respect to the control set point. From this graph, we can see the controller is able to control tube pressure to the set point.

The controller has several features that make it ideal for use in thermal oxidations. First, the controller is fail safe. If power fails, it will continue to control at the last programmed set point. If house exhaust fails, the valve inside the controller will open to its maximum position until sufficient exhaust is available for the controller to begin controlling again. Essentially, no matter which failure mode occurs, the user and the wafers are as safe as (or safer than) they would be without absolute pressure control.

A second feature of the controller is its ability to control the tube pressure in two different modes. In the first mode, the controller uses an absolute set point for process control. In the second, the controller normalizes the tube pressure to local barometric pressure for particle-free wafer loading and unloading. Without this second mode, the tube pressure would have to normalize itself when the door opened. If the process pressure were greater than atmospheric pressure, the particles disturbed by the pressure normalization would have the potential to deposit on the wafers as they were removed from the furnace. Under significantly negative pressure operation, when the door opened, particles would be swept in, depositing on the wafers before they even left the tube. To prevent both of these situations, the controller is able to balance the tube pressure with external barometric pressure on command, to create a slightly negative pressure in the tube. The slight negative pressure has a scavenging effect on the tube opening, minimizing the particle-generation risk of processing well outside the local barometric pressure range.

Different control modes

Martin Fallon proposes to run the oxidation process in a horizontal furnace at a slightly positive pressure with respect to the area's mean barometric pressure. Some vertical furnace manufacturers currently process slightly above atmospheric pressure as well. Other vertical furnace manufacturers, however, run the process slightly negative with respect to the local barometric pressure. The decision to run either a positive or negative pressure process comes with distinct benefits and drawbacks:

Positive pressure operation means the tube is pressurized with respect to the room. This ensures any leakage will be from the tube outward, reducing potential contamination to the wafers. Positive pressure operation is also easy to standardize across different altitudes, since the exhaust gases will travel easily from the positively pressurized tube into the negatively pressurized exhaust. Since growth rates are higher at higher pressures, operating at positive pressure will reduce the process times, increasing overall throughput. During extreme conditions, however, this mode of operation would pressurize the tube to as much as 13 in. of water (330 mm of water or 0.4 psi) above the cleanroom. There is some indication that 13 in. internal pressure is the limit of some furnace seals. Furnace seals may require improvement before positive pressure operation becomes a standard. However, because the storms that drive local barometric pressure down are occasional, the pressure on any given day is closer to the barometric maximum than to the minimum. This means the amount of time spent with the seals pressurized would be minimal.

As Martin Fallon points out, running a slightly positive pressure will benefit a horizontal furnace application for the following reasons:

• Higher pressure inside the tube can extend the life of the quartz, counteracting the effects of gravity at temperature.
• The gas exhaust is controlled so it exits via a plumbed route and not into the cleanroom.
• Quartz scavengers offer a higher level of cleanliness than stainless steel scavengers.

Negative pressure operation means the tube is under a slight vacuum with respect to the room. A high-integrity seal is important in this mode as well, as the process tube may run as much as 13 in. negative. Most furnaces have hardware already designed for negative tube pressure, so redesign of the seals may not be required. If negative pressure operation is to be maintained across all altitudes and climate changes, standard house exhaust may need to be supplemented by additional vacuum techniques, especially for sea-level operation. On a calm day at sea level, five in. of house exhaust pressure translates to an absolute pressure in the house exhaust of about 1020 mbar. If the selected pressure is 810 mbar, for example, the process tube will be significantly more negative than house exhaust. Some sort of boost will be required to extract the exhaust gases from the tube. Because lower pressures result in lower growth rates, operating in a negative pressure regime will result in longer process times.

Intermediate pressure operation means the tube may be positively or negatively pressurized with respect to the room. In this mode of operation, a process pressure is selected that is in between the high and low barometric pressures for the locations in which the process is to be run. Whether the tube is pressurized or under vacuum will depend on the location and the weather at any given time.

This control mode has many of the drawbacks of both positive and negative pressure operation. The tube seals must withstand both positive and negative pressurization. Leak integrity will be critical for oxide quality and to ensure containment of HCl within the process tool. When the tube is pressurized, the exhaust gases will flow freely into the house exhaust. During fair days at sea level, however, a boost to house vacuum may be necessary to maintain the selected pressure.

If the set point is at the midpoint of local barometric fluctuations, the system typically will run negative. Under these conditions, during extreme excursions, the pressure on the furnace seals will be half of what it would be in either positive or negative operation. Intermediate pressure operation may minimize many of the drawbacks for operation across a wide range of barometric pressures.

Testing to date

Initial process testing of the new control system supports its use as a buffer against changes in barometric pressure. Current process testing has controlled with a minor change in barometric pressure of 3 Torr (3 mm mercury).

Engineers from SVG (San Jose, Calif.) and Progressive Technologies (Tewksbury, Mass.) performed a series of experiments at SVG's facility in San Jose. We used an 800° C, 30-minute wet oxidation process in an SVG model 310, 300 mm Rapid Vertical Processor to evaluate the controller. We performed 16 runs with 200 mm wafers — the first eight runs without the controller, and the last eight runs with the controller set to 1041 mbar (780 Torr).

6. Absolute pressure control isolates the process from changes in house exhaust draw. The exhaust fluctuations shown are severe.

To simulate changes in atmospheric pressure, we ran the exhaust line in two configurations: one through the house exhaust system (a vacuum situation) and the second with the exhaust disconnected (there are no hazardous gases in the exhaust from this wet oxidation process). We ran the first two runs of each set of eight with the exhaust connected to house exhaust. We disconnected the exhaust from the house system for the next two runs. Then these four runs were repeated. Figure 6 shows the average run thickness changes that resulted when the house exhaust was disconnected and reconnected. The chart on the left shows the changes without the new control system; the 3s deviation was equivalent to 5.1% of the average thickness. The chart chart on the right shows the changes with the new controller — the 3s deviation was only 0.7% of the average thickness. This shows how the controller minimized the disturbance from the change in exhaust configuration.

We also tested the controller to ensure its compatibility with HCl. We performed two, back-to-back HCl steam clean cycles. During the clean cycles, we set the controller to control at 1041 mbar. Each cycle included a two-hour HCl/O₂ clean followed by a two-hour steam clean at 1050°C. The total time at 1050°C was eight hours. During this strenuous testing, no malfunctions occurred. When we inspected the valve, we found no indication of corrosion or excessive heat.

The next step

While we have shown the new controller has the capability to control to a set absolute pressure and minimize disturbance due to house exhaust changes, what we are most interested in is the ability of the controller to isolate the process tube from changes in barometric pressure. We plan to install the controller on a similar SVG furnace in a geographic location with more severe barometric swings for further evaluation. Based on the data provided by SC300 and Martin Fallon, these tests should show a significant increase in process Cpk as one more process variable is brought under control. •

How Process Chamber Pressure Control Works

The SENTRY 1510, made by Progressive Technologies (Tewksbury, Mass.), is a piston-based mechanical pressure regulator with electronic set point control. As shown in the figure, the downward force of gravity (w) and the upward force of the spring tension (tS) on the piston give it an effective weight (wE): wE = tS - w Two additional forces act on the piston: the cleanroom pressure pushing down and the process pressure pushing up. When a change occurs, due to a change in gas flow rate or house exhaust, the net force on the piston becomes unbalanced. The piston responds immediately (within 50 msec) by moving in the direction of the new net force, reestablishing balance. In a diffusion process tube where gas flow rates change and house exhaust is fluctuating constantly, the SENTRY 1510 prevents the overshoot and undershoot typically found in traditional electronic methods of process control. 1. The piston is a “fluidic calculator,” using physics to control the process pressure. While the piston mechanically maintains dynamic control, the microprocessor-based SENTRY Supervisor electronically controls the process set point. The Supervisor applies a PID algorithm to the sensed pressure and commands the stepper motor accordingly. The stepper motor adjusts the spring tension, changing the process pressure at which the piston maintains equilibrium. Installed at the process controller, the Supervisor allows the operator to view a continuous, 60-second graphic display of process pressure or exhaust flow, and to make real-time adjustments, if necessary.

Dan Hall, director of R&D for Progressive Technologies, Inc. has been a driving force in the company since 1992 and currently spearheads new product development. He has 16 years of design and development experience in the semiconductor, defense and industrial control industries. Dan holds a BSEE from the University of Delaware and an MSEE from the University of Michigan. Progressive Technologies designs and manufactures two-stage regulators and automatic fab balancing systems for the semiconductor and other cleanroom industries.Contact Hall at Progressive Technologies, Inc., 200 Ames Pond Drive, Tewksbury, MA 01876.
Phone: 1-978-863-1000
Fax: 1-978-863-1099
e-mail: dhall@progres.com

Cole Porter, applications lab manager for SVG’s Thermal Systems Division, leads the applications group in batch furnace product demonstration, process development and product improvement. He has 19 years of experience in the semiconductor industry. He has worked in the SVG process engineering group for 10 years focusing on new product technology enhancements and, most recently, the 300mm program. He holds three patents for SVG furnaces.Contact Porter at SVG, 2240 Ringwood Ave, San Jose, CA 95131.
Phone 1-408-944-8744.
Fax: 1-408-325-8898
e-mail: cporter@svg.com

Lise Laurin began her career as a process engineer at Intel and has held numerous positions in semiconductor processing and marketing over the last 19 years. She founded Clear Tech in 1996 to provide technical marketing and consulting services to the semiconductor supplier community. She has a B.S. in physics from Yale University, and is an active member of the Semiconductor Safety Association and the SEMI New England Committee.Contact Laurin at Clear Tech, 14 S. Main St., Newton, NH 03858
Phone: 1-603-382-7682
Fax: 1-603-382-0491
e-mail: llaurin@greennet.net

---

REFERENCES

1. Fallon, Martin, "Pressure Fix for Furnace Control," European Semiconductor, September 1999, pp 37-39.
   
2. Encyclopaedia Britannica, University of Chicago, (c)1990, Volume 1, p 676.
   
3. Ibid.
   
4. An unregulated process will have greater fluctuations due to changes in house exhaust draw caused by localized winds, cross-talk, etc. See Schanzer, et al, "Optimizing Thickness Uniformity with Furnace Pressure Control," Solid State Technology, June 1996, pp 127-133.
   

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS